html5-img
1 / 19

A multi-scale approach to InSAR time series analysis

M. Simons, E. Hetland, P. Muse, Y. N. Lin & C. DiCaprio. A multi-scale approach to InSAR time series analysis. Interferogram stack. A geophysical perspective on deformation tomography Examples: Long Valley Caldera, California Northern Volcanic Zone, Iceland. U. time.

dara
Download Presentation

A multi-scale approach to InSAR time series analysis

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. M. Simons, E. Hetland, P. Muse, Y. N. Lin & C. DiCaprio A multi-scale approach to InSAR time series analysis Interferogram stack A geophysical perspective on deformation tomography Examples: Long Valley Caldera, California Northern Volcanic Zone, Iceland U time

  2. Assume that in the future (Sentinel, DESDynI) we will have: • Frequent repeats (short DT) • Good orbits with small baselines • Ubiquitous high coherence Challenge for the future: • How to deal with O(103) interferograms • How to use Cd - Invert all pixels simultaneously? • 1000 igramsx 1000 x 1000 pixels = 1 billion data • Computational tractability • Approach (MInTS = Multi-scale InSAR Time Series): • Time domain: A generalized physical parameterization (GPS-like) • Space domain: Wavelets – use all data simultaneously • Note: Currently only applied to unwrapped data (we want this to change) Motivation

  3. How to deal with holes from decorrelation & unwrapping in individual scenes? We want to avoid the union of all holes in final time series. Interpolate in space? Interpolate in time? Our approach: Space & time simultaneously (tomography-like)

  4. MInTS Recipe • Interpolate unwrapping holes in each interferogram where needed (temporary) • Wavelet decomposition of each interferogram • For later weighting purposes, track relative extent to which each wavelet coefficient is associated with actual data versus interpolated data • Time series analysis on wavelet coefficients • Physical parameterization + splines for unknown signals - all constrained by weighted wavelet coefficients of observed interferograms • Recombine to get total deformation history

  5. Begin with a time domain synthetic for a “single pixel” Add noise

  6. InSAR time series: SBAS-pure (2 examples) Goal: Continuous displacement record Assumption: Constant Vi between image acquisitions dates Approach: Minimize Vi Method: SVD Secular + Seasonal + Transient + Noise + EQ + relaxation Secular + Seasonal + Transient + Noise

  7. Issues • Temporal parameterization controlled by dates of image • No explicit separation of processes contributing to the phase • Inherently pixel-by-pixel approach (computational restriction) • Ignores apriori information and data/model covariances

  8. MInTSGeneralized temporal description(geophysical intuition) Observation equation for LOS deformation: offsets (e.g., eqs., “fast” events) postseismic-like processes (i.e., one sided) periodic deformation (seasonal) generalized functions: B-splines, uniform (symmetric) and non-uniform (asymmetric, localized)

  9. Typical splines (current examples too coarse in time…) Regularization parameter, l, on B-splines chosen by cross-validation and is currently assumed constant for all wavelet scales and positions.

  10. Observation equation other “noise” (pixel & igram specific) orbital errors (igram specific) interferometric measurement orbital errors approximated by solve via (constrained) least squares – (generalized implementation): • optional constraints: • minimum soln. • flat soln. • smooth soln. • sparse soln. applied individually on each model parameter, or function Use cross-validation to determine l’s

  11. Demo time series – single pixel: Time series is parameterized physically - not with image acquisition times - disconnected sets do not require any additional parameters

  12. Now the spatial problem Neighboring pixels are highly correlated Hanssen, 2001 Emardson et al., 2003 Lohman & Simons, 2005 Challenge: Coupling all pixels together is expensive

  13. MInTS: Spatial parameterization Farras or Meyer 2D Wavelets 3 wavelets per scale (lossless)

  14. Wavelet parameterization permits implicit inclusion of Cd Cd= A exp(L/Lc) Lc= 25 km Corresponding wavelets coefficients nearly uncorrelated Note: wavelet approach has no master pixel – inherently relative displacements at a given scale

  15. Long Valley: (ERS/Envisat in geographic coordinates) – Velocity

  16. Example: Iceland Northern Volcanic Zone – Instantaneous Velocity

  17. Example: Iceland Northern Volcanic Zone – Instantaneous Velocity (nonlinear)

  18. Example: Iceland Northern Volcanic Zone – Instantaneous Velocity

  19. MInTS • Allows all scenes to be used even when isolated holes are present • Interpolates holes in time & space – “deformation tomography” • Implicitly includes expected data covariance • Physically parameterized but with ability to “discover” transients • Flexible choice of regularization on different components of the parameterization (smoothness, sparsity,…) • Computationally efficient • ----------------------- • Potential for incorporation of other data (e.g., GPS) and N,E,U parameterization given enough LOS diversity. • Potential for integrated phase unwrapping

More Related